This paper introduces a novel method for monitoring behavioral changes in university students by constructing spatiotemporal graphs from smartphone sensor data. Utilizing the Student Life dataset, which collects multi-modal data from smartphone sensors over a 10-week period, we capture detailed aspects of student behavior, including location, physical activity, and self-reported stress. By representing this data as spatiotemporal graphs, we model behavioral evolution across both temporal and spatial dimensions, employing a spatiotemporal Graph Neural Network (STGNN) to detect patterns associated with stress, sleep quality, and academic performance. This method enables a dynamic, high-resolution analysis of student well-being, offering a more comprehensive understanding of behavior over time.